Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

Health Services Research & Development

Veterans Crisis Line Badge
Go to the ORD website
Go to the QUERI website

RRP 12-447 – HSR&D Study

New | Current | Completed | DRA | DRE | Portfolios/Projects | Centers | Career Development Projects

RRP 12-447
Integrating Heart Failure Clinical Decision Support (CDS) at Point of Care
Mary K. Goldstein MD MS
VA Palo Alto Health Care System, Palo Alto, CA
Palo Alto, CA
Funding Period: January 2013 - December 2013

Clinical decision support (CDS) has the potential to assist health professionals managing patient care. Offering CDS within the context of the electronic health record (EHR) allows for patient-specific recommendations presented within existing clinical workflow, which is more efficient for providers than requiring access to a separate system for CDS. Current CDS within the EHR is typically limited to simple alerts and reminders, order sets, and documentation templates, which are highly useful tools, but do not harness the full power of CDS for complex patients. More extensive forms of CDS are available in systems external to the EHR. Linkages of external CDS to EHR are needed to bring the full potential of complex CDS to providers within their EHR workflow. In a previous VHA Strategic Innovations project, early work had been done on a Point of Care (POC) CDS method to interface an external CDS to CPRS, using the ATHENA-Hypertension (HTN) CDS as an example. That project developed POC CDS software that invoked the ATHENA-CDS for hypertension, exported data to it, and retrieved and formatted ATHENA-CDS recommendations for display as a dynamically-generated (rather than pre-coded) CPRS reminder. A dynamically-generated reminder allows for display of the specific items recommended for the patient under consideration, rather than a list of all possibilities across patients. It required further work for the methods developed in that project to apply to different clinical domains.
Heart failure (HF) is a prevalent condition with complex management choices that present CDS challenges that differ from those of hypertension. We had previously developed a CDS system for HF, known as ATHENA-HF, and we sought to develop a prototype linkage for HF to CPRS using the POC CDS approach.

The specific aims of this project were to: (1) establish health information technology (HIT) to interoperate the external CDS ATHENA-HF with CPRS using the new tools developed in the POC CDS project; (2) evaluate the accuracy of the recommendations generated by the new system using sample (de-identified or simulated) VA patient data; and (3) evaluate the integrated system with stakeholders.

The overall method used was iterative cycles of design, implementation, and testing of software in a test environment with CPRS.

Specific Aim 1:
We analyzed the HF guideline recommendations encoded in ATHENA-HF to identify items to update in order to align with the 2013 American College of Cardiology/American Heart Association (ACC/AHA) guidelines for the management of heart failure, drawing on the clinical expertise of our research team to interpret the guidelines and translate them into computable formats. To integrate the updated ATHENA-HF CDS with CPRS, research team and contractors collaborated on changes to the POC CDS software that had been developed for the previous VHA Innovations project, adapting it to the needs of HF CDS by (1) updating the knowledge base (KB) used by the POC CDS software and (2) modifying the list of clinical concepts that the software retrieved from the data available in CPRS, including additional tables for information specific to HF.

Specific Aim 2:
To ensure that the KB was producing the expected recommendations, we performed unit testing of the recommendations encoded in the KB. Hypothetical est patient profiles to use as positive controls were created with the appropriate characteristics, diagnoses, and medications to trigger recommendations for each of the following therapies: ACE-inhibitor, Aldosterone Receptor Blocker, Beta-Blocker, Aldosterone Antagonist, Hydralazine and Isosorbide Dinitrates, Cardiac Resynchronization Therapy, and Implantable Cardioverter-defibrillator. Hypothetical test patient profiles to use as negative controls were also created. The negative control cases had strong contraindications to each of the therapies included in the project, to test that recommendations were correctly suppressed. We also verified that the CDS would issue an alert if a patient were taking a drug designated as harmful to a HF patient, such as ibuprofen, and that if a patient were taking a beta blocker other than a recommended beta blocker (or sotalol, which has specialized uses apart from HF management), a message would be communicated to substitute that beta blocker for one recommended by the guideline. We unit tested 68 distinct criteria for contraindications to the recommended therapies. To evaluate the clinical accuracy of the recommendations generated by the updated HF CDS, we prepared de-identified patient data for offline testing by a physician, and prepared a test environment capable of producing CDS recommendations based on the de-identified data to compare CDS output with that of the physician. We compared the recommendations produced by the CDS and the recommendations produced by the physician for the same set of pilot test patients. This offline testing by a physician is conducted to identify errors in the encoding (that is, system does not do what we intended it to do) and errors in the way we interpreted the guidelines (that is, system does what the non-clinical encoders had intended it to do, but requires correction for clinical reasons). We then made changes to the KB based on these results, and unit tested the changes.

Specific Aim 3:
Our team evaluated the prototype integrated system in the VHA Innovations sandbox environment.

We updated the recommendations encoded in ATHENA-HF CDS to align with the 2013 ACC/AHA guidelines for the management of heart failure, drawing on the clinical expertise of our research team to interpret the guidelines and translate them into computable formats. We also expanded and enhanced the knowledge encoded in the ATHENA-HF KB, which now includes a greater range of contraindications to our recommended therapies, and more nuanced action choices to account for special cases. The updated ATHENA-HF KB was successfully imported into the POC CDS software. The integrated system produces prototype dynamically-generated, patient-specific recommendations and displays them within CPRS as CPRS reminders with clickable action buttons within the test environment. The VHA sandbox has test patient data simulating CPRS data; however, this does not include ejection fraction (EF) nor HF symptoms which are essential to clinical decision-making in HF but are not structured data in VistA. In order to test the ATHENA-HF-POC system, we simulated EF and New York Heart Association (NYHA) symptom class data and developed methods to input these to the CDS using VA FileMan, which provides application utilities to define, enter, and retrieve information from computer files.

Developing methods to automate computation of patient-specific guideline recommendations for patients with HF and presenting options within the CPRS workflow to facilitate selection of these recommendations for clinical management has potential to advance the capability of VA HF management. By developing a prototype system, this project is a step toward that goal. The project is significant in using a method to link an external CDS system for patients with complex conditions to an existing EHR, using the CPRS reminders; this paves the way for future systems in other clinical domains for which CDS processing requires systems external to CPRS.

None at this time.

DRA: Cardiovascular Disease
DRE: Prognosis
Keywords: none
MeSH Terms: none

Questions about the HSR&D website? Email the Web Team.

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.